How many hours does construction lose to rework and information friction

If you are making the case for coordination support or a new tool, someone will reach for a number: construction teams lose fourteen hours a week. The figure comes from a real report. It is also a 2018 survey composite of three different activities, and it is often quoted as something it is not.

The published research is broader than that one survey. Surveys report what people estimate. Timesheet, interview and observation studies record what staff and crews did under stated conditions. Process studies track how long a drawing waits. Each answers a different question.

Our review did not establish a current, representative weekly avoidable-loss rate across virtual design and construction (VDC), mechanical, electrical and plumbing (MEP) coordination, detailing and fabrication roles. That is a finding about the evidence we reviewed, not proof that no such study exists. Below is what the evidence does support, and a question list for testing any statistic.

The evidence at a glance

Read each row across before quoting a figure; the last column matters as much as the number.

Surveys: what respondents estimated

Office and detailing work: recorded or reconstructed

Site work: observed

Process flow: elapsed time and estimates

What the 14.1 hours contains

PlanGrid and FMI asked 599 construction leaders how many hours a week they spent on six activities. Three were classed as non-optimal: 5.5 hours searching for project information, 4.7 hours resolving conflicts, and 3.9 hours dealing with mistakes and rework. Together they make 14.1.[1]

Three details change how the figure can be used:

• Rework is 3.9 of the 14.1 hours. The rest is searching and conflict resolution.[1]
• Conflict here means disagreements among project participants, not model clashes.[1]
• The often-quoted 35% is 14.1 of the 40.5 hours respondents allocated across the six activities, not a share of a measured work week.[1]

The 19 VDC managers in the sample reported 14.42 hours.[1] A group of 19 cannot stand in for VDC managers in general.

Three surveys, three questions

Deloitte Access Economics' 2023 report for Autodesk found that management and executive employees spent 11.5 hours a week researching and analyzing data. The report calls an unquantified part of that time inefficient; it does not call the whole 11.5 hours lost.[2] Analysis is often the job.

Autodesk's 2025 report asked respondents about their own weekly time looking for data and reported a mean of 13 hours. Digitally advanced respondents reported two fewer hours.[3] That is a difference between groups, not the measured effect of buying software.

Different questions, samples and years mean these surveys cannot be averaged into one range. A move from 5.5 hours searching in 2018 to 13 hours looking for data in 2025 is not a trend either. Each figure belongs in a sentence with its own question, population and year.

Office and detailing work has been measured

It is sometimes said that nobody has measured detailers or design staff. Earlier drafts of our own review made that mistake.

Sacks and Barak drew engineering, detailing and drafting hours from staff timesheets on 52 precast and 14 cast-in-place projects, then compared them with modeling experiments. The authors note differences in skills and conditions between the office baseline and the experiments, and they excluded design-change hours from the comparison.[4]

Josephson and Saukkoriipi followed five architects and five technical consultants for one working day each, noting activity every two minutes. They recorded interruptions, waiting and some rework.[5]

Yoo and Ham compared 2D and 3D fabrication documentation on one plant-facility steel project. Working hours were reconstructed through interviews with engineers who did the work, not timed task by task, and repeated design changes were excluded.[6]

These studies record real work. None supplies a current weekly rate of avoidable loss for a coordinator, detailer or fabrication administrator.

Site studies separate support from waste

Josephson's defect research placed trained observers on seven Swedish building projects. They followed 2,879 defects and reported 22,000 correction hours, 7.1% of the labor hours worked in the study periods. The cost assessment included estimates.[7]

Newer observation shows why time outside direct work cannot simply be called waste. In Seppänen and Görsch's helmet-camera study, direct installation was a minority of analyzed time. The rest included discussion, hauling, searching and preparation. The authors traced most of the waste they did identify to missing prerequisites, such as materials, unfinished preceding work and missing design information.[8]

The sample was selected by availability during pandemic restrictions. The authors also note that their direct-work shares are lower than work sampling usually reports, and that the difference in method may explain part of the gap.[8]

Holth, Fosse and Drevland's five days of work sampling on one carpentry crew found that indirect work, mainly preparation and transport, was the largest category and wasted work the smallest.[9] It is one project, not a population estimate.

A waiting drawing is not an idle person

Flow studies measure how long work takes to move, not how long people sit still.

In Freire and Alarcón's study of four projects at one design firm, drawings were being worked on for 31.7% of their cycle time and spent the rest waiting between stages.[10] During that wait, the designer may well have been working on another drawing.

Chin's study logged 29 engineering reviews of shop drawings at one engineering firm over about seven months. The mean was 55.82 elapsed hours from stamp-in to stamp-out. The paper presents it as 6.98 days by dividing by eight, so it is neither seven calendar days nor reviewer labor. The author concluded that the root cause of long reviews was insufficient and unclear information rather than reviewer capability or availability.[11]

Nath and colleagues estimated a 36% processing-time and 38% total-time improvement for a re-engineered precast shop-drawing process. Parts were validated in a workshop and a pilot, the future workflow was estimated by experienced precasters, and whole-workflow validation was outside the study.[12] It is a labeled estimate, not an observed saving.

Test a statistic before it goes into a business case

The questions below are Atlas's checklist, drawn from the distinctions above. Apply them to any figure, including the ones in this article.

1. Whose hours: which roles, trades, country and year?
2. How were they obtained: self-estimate, timesheet, observation, interview reconstruction or projection?
3. Does one total combine several different activities?
4. What is the denominator: a paid week, a drawing's cycle time, project labor hours or cost?
5. Is the time necessary work, such as review, coordination, legitimate search or an owner-directed change?
6. How many people or items sit behind the figure, and is any subgroup small?
7. Who sponsored the study, and do they sell something related to the result?
8. Did the study measure a change, compare groups, or estimate a future workflow?
9. Does using it as weekly hours require an assumption you would have to state?
10. Would the released time become cash, or only capacity for other work?
11. Can a reviewer open the source and find the page?

Use the research to decide what to measure